Context

• This is an example of a notebook to demonstrate pycaret library for anomoly/outlier detection. In this example we will do some exploratory data analysis on the famous Iris dataset and then we build few models using pycaret for anomoly/outlier detection using supervised target feature.
• PyCaret is a Python open source machine learning library designed to make performing standard tasks in a machine learning project easy.
  • It is a flexible and low-code library that enhances productivity thus saving time and effort.
  • PyCaret is a simple and easy-to-use machine learning library that enables us to perform ML tasks within minutes.
  • PyCaret library allows automating the machine learning steps such as data transformation, preparation, hyper-parameter tuning, and standard model comparisons.

Data Attributes

• The Iris Dataset contains four features (length and width of sepals and petals) of 50 samples of three species of Iris (Iris setosa, Iris virginica and Iris versicolor).

# import libraries

import pandas as pd
import numpy as np
import seaborn as sns
import seaborn; seaborn.set()
import matplotlib.pyplot as plt
%matplotlib inline
import plotly.express as px
sns.set_theme()

import plotly.graph_objs as go
from plotly.offline import init_notebook_mode, iplot

init_notebook_mode(connected=True)  

import plotly.offline as pyo
import plotly.graph_objs as go
# Set notebook mode to work in offline
pyo.init_notebook_mode()

# import file
df=pd.read_csv('iris.csv')

# shape of the data
df.shape

(150, 5)

# get basic info
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 150 entries, 0 to 149
Data columns (total 5 columns):
 #   Column        Non-Null Count  Dtype  
---  ------        --------------  -----  
 0   sepal_length  150 non-null    float64
 1   sepal_width   150 non-null    float64
 2   petal_length  150 non-null    float64
 3   petal_width   150 non-null    float64
 4   target        150 non-null    int64  
dtypes: float64(4), int64(1)
memory usage: 6.0 KB

# distribution of the data
plt.figure(figsize=(15,6))
sns.boxplot(x="variable", y="value", showmeans=True, data=pd.melt(df))
plt.show()

# relationship between variables.
sns.pairplot(df);

• We see linear relationship among the variables.
• We can indentify 2 clusters.

Anomoly/Outlier Detection model building.

from pycaret.anomaly import *
setup = setup(df, session_id = 123)

	Description	Value
0	session_id	123
1	Original Data	(150, 5)
2	Missing Values	False
3	Numeric Features	4
4	Categorical Features	1
5	Ordinal Features	False
6	High Cardinality Features	False
7	High Cardinality Method	None
8	Transformed Data	(150, 7)
9	CPU Jobs	-1
10	Use GPU	False
11	Log Experiment	False
12	Experiment Name	anomaly-default-name
13	USI	0f14
14	Imputation Type	simple
15	Iterative Imputation Iteration	None
16	Numeric Imputer	mean
17	Iterative Imputation Numeric Model	None
18	Categorical Imputer	mode
19	Iterative Imputation Categorical Model	None
20	Unknown Categoricals Handling	least_frequent
21	Normalize	False
22	Normalize Method	None
23	Transformation	False
24	Transformation Method	None
25	PCA	False
26	PCA Method	None
27	PCA Components	None
28	Ignore Low Variance	False
29	Combine Rare Levels	False
30	Rare Level Threshold	None
31	Numeric Binning	False
32	Remove Outliers	False
33	Outliers Threshold	None
34	Remove Multicollinearity	False
35	Multicollinearity Threshold	None
36	Remove Perfect Collinearity	False
37	Clustering	False
38	Clustering Iteration	None
39	Polynomial Features	False
40	Polynomial Degree	None
41	Trignometry Features	False
42	Polynomial Threshold	None
43	Group Features	False
44	Feature Selection	False
45	Feature Selection Method	classic
46	Features Selection Threshold	None
47	Feature Interaction	False
48	Feature Ratio	False
49	Interaction Threshold	None

• Specify the session id, this results in processing after execution. It interprets numerous types of variables automatically and allows us to confirm by pressing ENTER to continue.

• Observe that our dataset consists of 5, 150 rows each. We can perform various imputations- numeric and categorical or normalize the data. But we don’t require such transformations in our dataset so let us continue!

• Performing all these computations with a few lines of code presents the beauty of the PyCaret library.

# get list of models.
models()

	Name	Reference
ID
abod	Angle-base Outlier Detection	pyod.models.abod.ABOD
cluster	Clustering-Based Local Outlier	pyod.models.cblof.CBLOF
cof	Connectivity-Based Local Outlier	pyod.models.cof.COF
iforest	Isolation Forest	pyod.models.iforest.IForest
histogram	Histogram-based Outlier Detection	pyod.models.hbos.HBOS
knn	K-Nearest Neighbors Detector	pyod.models.knn.KNN
lof	Local Outlier Factor	pyod.models.lof.LOF
svm	One-class SVM detector	pyod.models.ocsvm.OCSVM
pca	Principal Component Analysis	pyod.models.pca.PCA
mcd	Minimum Covariance Determinant	pyod.models.mcd.MCD
sod	Subspace Outlier Detection	pyod.models.sod.SOD
sos	Stochastic Outlier Selection	pyod.models.sos.SOS

• List of model can be build from the above table for Anomoly/Outlier detection.
• For this excercise I will build 3 models (iforest, loc and Outlier).

iForest model

# building iforest model with estimator as random forest.
iforest_model = tune_model(model = 'iforest', supervised_target ='target', supervised_estimator='rf')

	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC
0.0	0.9467	0.0000	0.3800	0.4000	0.3894	nan	0.2000
0.01	0.9452	0.0000	0.3800	0.4000	0.3894	nan	0.2000
0.02	0.9443	0.0000	0.3800	0.4000	0.3894	nan	0.2000
0.03	0.9433	0.0000	0.3790	0.4000	0.3889	nan	0.2000
0.04	0.9429	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.05	0.9429	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.06	0.9357	0.0000	0.3714	0.4000	0.3846	nan	0.2000
0.07	0.9429	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.08	0.9484	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.09	0.9478	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.1	0.9467	0.0000	0.3775	0.4000	0.3880	nan	0.2000

• From the above we can observe that when the when the anomoly fraction is at 0 the accuracy of the model was at 0.9467 and then model accuracy gradually drops, after 0.06 anomoly fraction we can can see the model accuracy incrases and the final model accuracy is at 0.9467.

Local outlier factor model building

# building loc model with estimator as random forest.
lof_model = tune_model(model = 'lof', supervised_target ='target', supervised_estimator='rf')

	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC
0.0	0.9467	0.0000	0.3800	0.4000	0.3894	nan	0.2000
0.01	0.9386	0.0000	0.3733	0.3889	0.3798	nan	0.1873
0.02	0.9381	0.0000	0.3733	0.3889	0.3798	nan	0.1873
0.03	0.9367	0.0000	0.3724	0.4000	0.3852	nan	0.2000
0.04	0.9429	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.05	0.9429	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.06	0.9429	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.07	0.9429	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.08	0.9412	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.09	0.9401	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.1	0.9319	0.0000	0.3703	0.4000	0.3840	nan	0.2000

• Local Outlier Factor model accuracy is at 0.9319.

Cluster model building

# building cluster model with estimator as random forest.
cluster_model = tune_model(model = 'cluster', supervised_target ='target', supervised_estimator='rf')

	Accuracy	AUC	Recall	Prec.	F1	Kappa	MCC
0.0	0.9467	0.0000	0.3800	0.4000	0.3894	nan	0.2000
0.01	0.9452	0.0000	0.3800	0.3889	0.3835	nan	0.1873
0.02	0.9448	0.0000	0.3800	0.3875	0.3827	nan	0.1875
0.03	0.9367	0.0000	0.3724	0.4000	0.3852	nan	0.2000
0.04	0.9357	0.0000	0.3714	0.4000	0.3846	nan	0.2000
0.05	0.9500	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.06	0.9429	0.0000	0.3714	0.4000	0.3846	nan	0.2000
0.07	0.9500	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.08	0.9484	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.09	0.9478	0.0000	0.3786	0.4000	0.3886	nan	0.2000
0.1	0.9544	0.0000	0.3775	0.4000	0.3880	nan	0.2000

• From the above cluster model we can observe that the accuracy is at 0.9544
• From the three models I have built I can obseve that the Cluster model having higher accuracy.

# assign the cluster model to data frame.
cluster_results = assign_model(cluster_model)
cluster_results.head()

	sepal_length	sepal_width	petal_length	petal_width	target	Anomaly	Anomaly_Score
0	5.1	3.5	1.4	0.2	0	0	0.268103
1	4.9	3.0	1.4	0.2	0	0	0.223945
2	4.7	3.2	1.3	0.2	0	0	0.141955
3	4.6	3.1	1.5	0.2	0	0	0.141955
4	5.0	3.6	1.4	0.2	0	0	0.298188

# frequencies of anomolies in the cluster model.
cluster_results.Anomaly.value_counts(normalize=True)

0    0.9
1    0.1
Name: Anomaly, dtype: float64

• Cluster model recognises that 1% of the data points are anomolies.

# 2d view of Anomoly
evaluate_model(cluster_model);

• Highlighed yellow circles are the anomolies.

# scatter plot
evaluate_model(cluster_model);

• We can zoom this 2D plot to view which points were considered as outliers/anomolies.

# lets creeate an array for inliner and outlier

inliner=np.array(cluster_results.loc[(cluster_results["Anomaly"])==0]["Anomaly_Score"])
outlier=np.array(cluster_results.loc[(cluster_results["Anomaly"])==1] ["Anomaly_Score"])

# create a histrogram for inliners and outliers.

x=inliner
y=outlier
plt.figure(figsize=(10,4))

plt.hist(x,alpha=0.5,label='inliner')
plt.hist(y,alpha=0.5, label='Outlier')
plt.legend(loc='upper left')
plt.show()

• We can observe above that the outliers having higher anomoly score >0.75.

# pairplot of the cluster model with anomoly as hue
sns.pairplot(cluster_results, hue = "Anomaly");

• From the above we can obseve how the anomolies scattered across.